Welcome Guest! To enable all features please Login or Register.

Notification

Icon
Error

Login


2 Pages12>
Options
View
Go to last post Go to first unread
miero  
#1 Posted : Saturday, February 15, 2014 11:04:43 PM(UTC)
miero

Rank: Member

Groups: Member
Joined: 10/22/2011(UTC)
Posts: 135
Location: Prague

Thanks: 1 times
Was thanked: 3 time(s) in 3 post(s)
I've managed to connect I2S output of RPi to B3 successfully.

Although it plays very well, it bothers me that the output with I2S so not so clean as it is with TOSLINK.

Using my limited resources I've tried to measure waveform of small signal and also spectrum of playing digital silence. I'm afraid that that's limit I can do.

Here is my setup:
- Buffalo 3 + IVY
- 100MHz scope and probe
- Raspberry Pi powered by switched power adapter (not connected to earth)
- RPi connected to B3 via 7cm cable, shielded by aluminum foil, "grounded" to GND
- RPi controlled by Ethernet; i.e. no USB or HDMI
- Toslink module connected via 15cm cable, shielded like I2S cable
- 30cm RCA cable, isolated from the chasis
- Output of DAC floating, grounded only via scope probe
- RPi and TOSLINK connected to B3 at the same time
- TOSLINK at input #3; optical cable disconnected when I2S plays

My suspicions:
1) RPi power supply is dirty; noise leaks to B3 when I2S is transmitting
2) I2S path from RPi to B3 is not well shielded; it radiates and induces at other place; better if it would be whole at one PCB
3) ???

Is my measurement showing something? Can it be improved? I've tried 3 switching power adapters for RPi and each one shown a little bit different HF/VHF noise. Maybe a good linear supply might help.

Thanks for comments.
miero attached the following image(s):
Buffalo3_IVY_Toslink_vs_RPiI2S.png (22kb) downloaded 71 time(s).

You cannot view/download attachments. Try to login or register.
Russ White  
#2 Posted : Saturday, February 15, 2014 11:16:45 PM(UTC)
Russ White

Rank: Administration

Groups: Administration, Customer
Joined: 10/24/2006(UTC)
Posts: 3,979
Location: Nashville, TN

Thanks: 25 times
Was thanked: 89 time(s) in 83 post(s)
Just curious, what sample rate are you using? Is any sample rate conversion happening?
Russ White  
#3 Posted : Saturday, February 15, 2014 11:22:32 PM(UTC)
Russ White

Rank: Administration

Groups: Administration, Customer
Joined: 10/24/2006(UTC)
Posts: 3,979
Location: Nashville, TN

Thanks: 25 times
Was thanked: 89 time(s) in 83 post(s)
Teleporters could also solve the issue.
Russ White  
#4 Posted : Saturday, February 15, 2014 11:42:36 PM(UTC)
Russ White

Rank: Administration

Groups: Administration, Customer
Joined: 10/24/2006(UTC)
Posts: 3,979
Location: Nashville, TN

Thanks: 25 times
Was thanked: 89 time(s) in 83 post(s)
Any chance you could post pictures of your test rig? It could be simple wire routing could also help a lot.
DQ828  
#5 Posted : Sunday, February 16, 2014 3:54:45 AM(UTC)
DQ828

Rank: Member

Groups: Member
Joined: 8/17/2010(UTC)
Posts: 368
Location: australia

Thanks: 8 times
Was thanked: 3 time(s) in 3 post(s)
For what it's worth http://www.crazy-audio.c...urces-on-a-raspberry-pi/

http://www.crazy-audio.c...y-noise-on-the-5v-rails/

Edited by user Sunday, February 16, 2014 3:55:36 AM(UTC)  | Reason: Not specified

Russ White  
#6 Posted : Sunday, February 16, 2014 4:19:01 AM(UTC)
Russ White

Rank: Administration

Groups: Administration, Customer
Joined: 10/24/2006(UTC)
Posts: 3,979
Location: Nashville, TN

Thanks: 25 times
Was thanked: 89 time(s) in 83 post(s)
That is interesting stuff. I have been wireless - and I am definitely not seeing anything like Miero.

I am also using a linear regulator, not a switching one.
miero  
#7 Posted : Sunday, February 16, 2014 11:50:05 AM(UTC)
miero

Rank: Member

Groups: Member
Joined: 10/22/2011(UTC)
Posts: 135
Location: Prague

Thanks: 1 times
Was thanked: 3 time(s) in 3 post(s)
The data used for measurements are generated live on RPi by sox tool, so there are not interrupts
from Ethernet or Wifi.

Update:

There was a lot of EM energy radiated from RPi USB cable from switching power supply, so
decided to replace it by linear based on ST L7805 regulator.

The power/ground line gets polluted by "I2S noise" too when RPi is playing.

Notice: My test rig is grounded only via oscilloscope ground channel, so I'm measuring things that usually goes to ground. So this might be reason why I'm getting different results.

My test rig:
- http://mr.ieero.com/pub/...e/i2s_noise_test_rig.jpg
- http://mr.ieero.com/pub/...oise_test_rig_detail.jpg

Attaching more measurements of noise on power line and also on the audio output with the linear regulator for RPi.

Next picture shows difference when playing silence between 96kHz and 192kHz noise.

Edited by user Sunday, February 16, 2014 11:54:39 AM(UTC)  | Reason: Not specified

miero attached the following image(s):
noise_power_vs_output_desc.png (13kb) downloaded 21 time(s).
audioout_noise_i2s_silent_both_desc.png (6kb) downloaded 10 time(s).

You cannot view/download attachments. Try to login or register.
Russ White  
#8 Posted : Sunday, February 16, 2014 1:27:57 PM(UTC)
Russ White

Rank: Administration

Groups: Administration, Customer
Joined: 10/24/2006(UTC)
Posts: 3,979
Location: Nashville, TN

Thanks: 25 times
Was thanked: 89 time(s) in 83 post(s)
It is not interrupts we are concerned about, but noise. :) Try it with wireless instead of wired ethernet.

I am very curious about your input wiring... That looks a bit suspect. You would be a lot better off to use the jumpers on the B3 to tie the PCM data lines together.

it is connected to a data input and if you are powering your toslink module it will be driven by it (even with no music playing - it will be either high or low) this is not good. :)

If you want to use some spdif source along with I2S you would be better to physically switch it out when not in use.


Russ White  
#9 Posted : Sunday, February 16, 2014 1:54:20 PM(UTC)
Russ White

Rank: Administration

Groups: Administration, Customer
Joined: 10/24/2006(UTC)
Posts: 3,979
Location: Nashville, TN

Thanks: 25 times
Was thanked: 89 time(s) in 83 post(s)
One could also try a bypass and decoupling capacitor at the 5V and GND pins of P5 on the pi.
miero  
#10 Posted : Sunday, February 16, 2014 2:12:45 PM(UTC)
miero

Rank: Member

Groups: Member
Joined: 10/22/2011(UTC)
Posts: 135
Location: Prague

Thanks: 1 times
Was thanked: 3 time(s) in 3 post(s)
There is no difference if Ethernet is used or not. I've tried to disconnect it completely and nothing has changed.

Also the same situation is with Toslink module. I've tried to disconnect it from the power and also no change. So I've kept it.

RPi is connected to B3 via 4 wires. B3 has installed 0Ohm jumpers for "smart wiring" for combined I2S/DSD playback. Check the picture: http://mr.ieero.com/pub/...e/b3_i2s_dsd_jumpers.jpg

I'd say there is problem with RPi because it generates much noise on I2S line which is then radiated on the way from RPi chip to DAC. It might not be issue if both chips were on the same PCB separated 1-2cm.
Russ White  
#11 Posted : Sunday, February 16, 2014 2:38:52 PM(UTC)
Russ White

Rank: Administration

Groups: Administration, Customer
Joined: 10/24/2006(UTC)
Posts: 3,979
Location: Nashville, TN

Thanks: 25 times
Was thanked: 89 time(s) in 83 post(s)
You should also think about terminating the I2S lines. Series termination would be a good start.

Also the bypass and decoupling (perhaps 220uf + 100nf) (I would try the 3.3V pin first actually) I mentioned should help greatly.

Also you need to establish some benchmark for your setup - another I2S source?

If you using the same scheme as B3SE/BII then you should be using D1 for SPDIF input, not D3.

Also for best results SPDIF detection should be turned off for I2S playback.

Also, once again, teleporters should solve the problem nicely - I think you are just seeing the effects of a long run with no termination and bad decoupling. They together can lead to some pretty high ground bounce at VHF - likely won't cause any audible issues, but it concerns you it can very definitely be fixed.

Edited by user Sunday, February 16, 2014 3:56:53 PM(UTC)  | Reason: Not specified

Russ White  
#12 Posted : Sunday, February 16, 2014 4:01:22 PM(UTC)
Russ White

Rank: Administration

Groups: Administration, Customer
Joined: 10/24/2006(UTC)
Posts: 3,979
Location: Nashville, TN

Thanks: 25 times
Was thanked: 89 time(s) in 83 post(s)
For what it's worth - I just tried 220uf + 100nf at 3.3V(pin 2) to GND (pin 8) on P5 of the pi. This did remove quite a lot of ground bounce at the pi. Also - it should be noted that I am using B3SE which adds some series termination because of the different input scheme. For B3 I would use 22-47R on each signal. It could also be beneficial to add a 100-220nf bypass at 5V. I didn't see much change when I did that though.

Edited by user Sunday, February 16, 2014 4:13:40 PM(UTC)  | Reason: Not specified

miero  
#13 Posted : Sunday, February 16, 2014 5:34:28 PM(UTC)
miero

Rank: Member

Groups: Member
Joined: 10/22/2011(UTC)
Posts: 135
Location: Prague

Thanks: 1 times
Was thanked: 3 time(s) in 3 post(s)
I'd like to report partial success.

I've shortened jumper cable to 5cm, added some caps I've found in the house (10uF + 100nF) on the 5V pin and also connected power from supply right there instead of USB connector.

This reduced noise above 100MHz and also bit at lower frequencies.

Now I can confirm that Ethernet adds noise and it could be better to use Wifi dongle.

After this tuning I've compared again the effect of linear power supply. It is better than switched PSU.

Although this helped a lot, the Toslink is still better.

Russ, can you recommend a linear PSU project for RPi? Or create TPA Low Voltage High Current one? ;-)

Thanks for help.

UserPostedImage
miero attached the following image(s):
aout_noise_regdiff_desc.png (7kb) downloaded 24 time(s).

You cannot view/download attachments. Try to login or register.
Russ White  
#14 Posted : Sunday, February 16, 2014 7:26:40 PM(UTC)
Russ White

Rank: Administration

Groups: Administration, Customer
Joined: 10/24/2006(UTC)
Posts: 3,979
Location: Nashville, TN

Thanks: 25 times
Was thanked: 89 time(s) in 83 post(s)
You can setup a LCDPS with a jumper at the CRC (instead of an R) and it can do up to 1A (which is all you should need) as long as you have a sufficient transformer. The GPIO of the pi run off the 3.3V supply, so you would be better to bypass those as well.

Teleporter was designed just for what you are trying to do. :)

In the end I think you've probably worried it about for more than it really matters. :)

Edited by user Sunday, February 16, 2014 7:36:40 PM(UTC)  | Reason: Not specified

Russ White  
#15 Posted : Sunday, February 16, 2014 7:29:39 PM(UTC)
Russ White

Rank: Administration

Groups: Administration, Customer
Joined: 10/24/2006(UTC)
Posts: 3,979
Location: Nashville, TN

Thanks: 25 times
Was thanked: 89 time(s) in 83 post(s)
Actually the newly redesigned Placid HD (beta) should also be able to do 1A at 5V as long as the input voltage does not exceed about 7.5 - 8V and you use sufficient size heatsinks.
miero  
#16 Posted : Sunday, February 16, 2014 9:43:12 PM(UTC)
miero

Rank: Member

Groups: Member
Joined: 10/22/2011(UTC)
Posts: 135
Location: Prague

Thanks: 1 times
Was thanked: 3 time(s) in 3 post(s)
Russ, thanks very much for your help.
Here are the result with caps added also on 3.3V and LCDPS with bridged R.
miero attached the following image(s):
before_after.png (29kb) downloaded 60 time(s).

You cannot view/download attachments. Try to login or register.
Russ White  
#17 Posted : Monday, February 17, 2014 11:06:42 PM(UTC)
Russ White

Rank: Administration

Groups: Administration, Customer
Joined: 10/24/2006(UTC)
Posts: 3,979
Location: Nashville, TN

Thanks: 25 times
Was thanked: 89 time(s) in 83 post(s)
Glad to help. :)

Teleporters.
miero  
#18 Posted : Tuesday, February 18, 2014 5:16:56 PM(UTC)
miero

Rank: Member

Groups: Member
Joined: 10/22/2011(UTC)
Posts: 135
Location: Prague

Thanks: 1 times
Was thanked: 3 time(s) in 3 post(s)
The RPi I had was borrowed, so Teleporters was not right solution for my temporary problem.

However, I've decided for BBB so I might try Teleporters later ;-)
DQ828  
#19 Posted : Wednesday, February 19, 2014 9:41:24 AM(UTC)
DQ828

Rank: Member

Groups: Member
Joined: 8/17/2010(UTC)
Posts: 368
Location: australia

Thanks: 8 times
Was thanked: 3 time(s) in 3 post(s)
Originally Posted by: Russ White Go to Quoted Post
Glad to help. :)

Teleporters.


How would the Teleporters be applied given the BIII & RPi would only be inches apart?

I'm assuming the signal is coming from the router via a wall outlet.

What if wireless was used?


miero  
#20 Posted : Wednesday, February 19, 2014 11:47:05 AM(UTC)
miero

Rank: Member

Groups: Member
Joined: 10/22/2011(UTC)
Posts: 135
Location: Prague

Thanks: 1 times
Was thanked: 3 time(s) in 3 post(s)
Originally Posted by: DQ828 Go to Quoted Post
What if wireless was used?


I have not tried, but the Wireless transmitter if close to RPi and DAC might/will be another noise source.

Regarding Teleporter, I wonder how noise spectrum it produces itself...
Rss Feed  Atom Feed
Users browsing this topic
GuestUser (3)
2 Pages12>
Forum Jump  
You cannot post new topics in this forum.
You cannot reply to topics in this forum.
You cannot delete your posts in this forum.
You cannot edit your posts in this forum.
You cannot create polls in this forum.
You cannot vote in polls in this forum.